Building Sustainability
Uine Kailamäki
2021
For the reader: You can click the black “code” box on the right hand side to either see or hide the R code behind analysis results and graphs.
The results are arranged to separate tabs by site/area, so that it’s easy to look at just the results of the site of interest, instead of scrolling accidentally down to a different site and getting confused.
The results for each site consist of four parts: pXRF, particle size distribution, loss on ignition, and colorimetry.
Sites
Ashdod-Yam
Intro about the site
pXRF
From literature we can identify three rough classes of elements that could be used to analyze mudbricks from pXRF results:
Good: Rb, Sr, Y, Zr, Nb, Th, Pb, Cu, Zn, Fe
Neutral: V, Cr, Co, Ni, Ca, Ti
Bad: Na, P, Ba (+ Ca, Al)
The “bad” elements were excluded from the pXRF results from the get-go. After scaling and averaging the remaining values we ended up with two analysis datasets: “pxrf_all” that contains all the elements that have non-zero results, and “pxrf_no_na”, in which all the elements that have even one NA result are omitted.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggfortify); # PCA autoplots
library(FactoMineR); # Fast PCA graphs
library(dendextend); # Dendrograms
# Read the raw data
pxrf <- read.xlsx("data/pxrf_AY.xlsx", sep=";")
# Remove any rows containing words 'ERROR' or 'TEST'
pxrf <- filter(pxrf, !grepl('TEST', Sample))
pxrf <- filter(pxrf, !grepl('ERROR', Sample))
# Remove unused columns
pxrf <- pxrf %>%
select(-Site) %>%
select(-'Other;name') %>%
select(-'File;#') %>%
select(-DateTime) %>%
select(-Application) %>%
select(-Method) %>%
select(-ElapsedTime) %>%
select(-'Cal;Check')
# Coercing "< LOD" -results to NA:s
pxrf[4:93] <- pxrf[4:93] %>% mutate_if(is.character,as.numeric)
# Averaging results by "Sample", "Area" and "Type" columns
pxrf_averaged <- aggregate(pxrf, by = list(pxrf$Sample, pxrf$Area, pxrf$Type), FUN = mean)
# Removing old (now empty) columns
pxrf_averaged <- pxrf_averaged %>% select(-Sample)
pxrf_averaged <- pxrf_averaged %>% select(-Area)
pxrf_averaged <- pxrf_averaged %>% select(-Type)
# Assigning sample names as new row names
rownames(pxrf_averaged) <- pxrf_averaged$Group.1
pxrf_averaged <- pxrf_averaged %>% select(-Group.1)
# Renaming the other two newly created groups back to "Type" and "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.2"] <- "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.3"] <- "Type"
# Removing error columns from the scaled analysis data set
pxrf_scaled <- pxrf_averaged %>% select(-ends_with("Err"))
# Scaling the data with standard z-score method
pxrf_scaled[3:47] <- scale(pxrf_scaled[3:47])
# Removing columns that only contain NA values
all_na <- function(x) any(!is.na(x))
pxrf_all <- pxrf_scaled %>% select_if(all_na)pxrf_averaged: All averaged data including averaged errors (Uine’s note: for later use)
colnames(pxrf_averaged)## [1] "Area" "Type" "MgO" "MgO;Err" "Al2O3" "Al2O3;Err"
## [7] "SiO2" "SiO2;Err" "P2O5" "P2O5;Err" "S" "S;Err"
## [13] "Cl" "Cl;Err" "K2O" "K2O;Err" "CaO" "CaO;Err"
## [19] "Ti" "Ti;Err" "V" "V;Err" "Cr" "Cr;Err"
## [25] "Mn" "Mn;Err" "Fe" "Fe;Err" "Co" "Co;Err"
## [31] "Ni" "Ni;Err" "Cu" "Cu;Err" "Zn" "Zn;Err"
## [37] "As" "As;Err" "Se" "Se;Err" "Rb" "Rb;Err"
## [43] "Sr" "Sr;Err" "Y" "Y;Err" "Zr" "Zr;Err"
## [49] "Nb" "Nb;Err" "Mo" "Mo;Err" "Rh" "Rh;Err"
## [55] "Pd" "Pd;Err" "Ag" "Ag;Err" "Cd" "Cd;Err"
## [61] "Sn" "Sn;Err" "Sb" "Sb;Err" "Ba" "Ba;Err"
## [67] "La" "La;Err" "Ce" "Ce;Err" "Hf" "Hf;Err"
## [73] "Ta" "Ta;Err" "W" "W;Err" "Pt" "Pt;Err"
## [79] "Au" "Au;Err" "Hg" "Hg;Err" "Tl" "Tl;Err"
## [85] "Pb" "Pb;Err" "Bi" "Bi;Err" "Th" "Th;Err"
## [91] "U" "U;Err"pxrf_all: All elements (the numbers below are the amount of NA values)
colSums(is.na(pxrf_all))## Area Type MgO Al2O3 SiO2 P2O5 S Cl K2O CaO Ti V Mn
## 0 0 1 0 0 0 43 1 0 0 0 28 0
## Fe Ni Cu Zn As Rb Sr Y Zr Nb Ba
## 0 1 0 0 27 0 0 2 0 31 45pxrf_final: Final analysis data set with only the selected elements
# Dropping the dubious elements and elements with too many NA values
pxrf_final <- pxrf_all %>%
select(-Al2O3) %>%
select(-CaO) %>%
select(-SiO2) %>%
select(-Cl) %>%
select(-K2O) %>%
select(-Ni) %>%
select(-P2O5) %>%
select(-S) %>%
select(-V) %>%
select(-As) %>%
select(-Nb) %>%
select(-Ba)
# Transforming remaining NA values to zeros for analysis
pxrf_final[is.na(pxrf_final)] <- 0
# Final analysis data set
colnames(pxrf_final)## [1] "Area" "Type" "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr"
## [11] "Y" "Zr"pXRF: K means cluster analysis
# PC autoplot with K-means clusters
set.seed(1)
autoplot(kmeans(pxrf_final[3:12], 7), data = pxrf_final, label = TRUE, label.size = 3)
pXRF: PCA
PCA with final elements:
# Elements
colnames(pxrf_final[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_1 <- prcomp(pxrf_final[3:12])
summary(pca_1)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.3836 1.2233 1.0086 0.84957 0.60483 0.49903 0.44478
## Proportion of Variance 0.5719 0.1506 0.1024 0.07265 0.03682 0.02507 0.01991
## Cumulative Proportion 0.5719 0.7225 0.8249 0.89752 0.93434 0.95941 0.97932
## PC8 PC9 PC10
## Standard deviation 0.3538 0.22250 0.17534
## Proportion of Variance 0.0126 0.00498 0.00309
## Cumulative Proportion 0.9919 0.99691 1.00000# PCA plots
biplot(pca_1, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Ashdod-Yam elements")
autoplot(pca_1, data=pxrf_final, colour='Area', shape = FALSE, label = TRUE, main = "PCA Ashdod-Yam grouped by area")
autoplot(pca_1, data=pxrf_final, colour='Type', shape = FALSE, label = TRUE, main = "PCA Ashdod-Yam grouped by sample type")
PCA(pxrf_final[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 47 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"pXRF: HCA
# HCA dendrogram, samples color coded by area:
dend <-
pxrf_final %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=7)
area <- as.factor(pxrf_final[, 1])
n_area <- length(unique(area))
cols_a <- colorspace::rainbow_hcl(n_area, c = 70, l = 50)
col_area <- cols_a[area]
labels_colors(dend) <- col_area[order.dendrogram(dend)]
plot(dend, main="HCA with sample areas")
rect.dendrogram(dend, 7, border = "Black")
legend("topright", legend = levels(area), fill = cols_a)
# HCA dendrogram, samples color coded by type:
dend2 <-
pxrf_final %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=7)
type <- as.factor(pxrf_final[, 2])
n_type <- length(unique(type))
cols_t <- colorspace::rainbow_hcl(n_type, c = 70, l = 50)
col_type <- cols_t[type]
labels_colors(dend2) <- col_type[order.dendrogram(dend2)]
plot(dend2, main="HCA with sample types")
rect.dendrogram(dend, 7, border = "Black")
legend("topright", legend = levels(type), fill = cols_t)
Grain size analysis
Data preparation
# Libraries
library(dplyr);
library(openxlsx);
library(ggtern)
# Read and filter data
grain <- read.xlsx("data/grain_AY.xlsx", sep=";")
# Remove automatically created averages in order to include results from multiple runs for the same sample
grain <- grain %>%
filter(!grepl('Average', Sample)) %>%
filter(!grepl('test', Sample))
# Average data by "Sample", "Context" and "Type" columns from the original dataset
grain <- aggregate(grain, by=list(grain$Sample, grain$Context, grain$Type), FUN=mean)
# Remove the now empty columns for clarity
grain <- grain %>% select(-Sample)
grain <- grain %>% select(-Context)
grain <- grain %>% select(-Type)
# Assign sample names ("Group.1") as row names, and rename the other created columns back to "Type" and "Area" for clarity
rownames(grain) <- grain$Group.1
grain <- grain %>% select(-Group.1)
names(grain)[names(grain) == "Group.2"] <- "Context"
names(grain)[names(grain) == "Group.3"] <- "Type"
# Scaling clay-silt-sand portions to values between 0 and 1
# (Otherwise the differences in low clay percentages are too small to differentiate)
normalize <- function(x, na.rm=TRUE){(x-min(x, na.rm=TRUE))/(max(x, na.rm=TRUE)-min(x, na.rm=TRUE))}
# Apply function "normalize" to clay, silt and sand columns
grain_01 <- as.data.frame(apply(grain[4:6], 2, normalize))Ternary plots
# Ternary plots
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=0) +
geom_text(aes(label=rownames(grain_01)), size=3)
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=2)
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Type)) +
labs(title="Clay silt sand") +
theme_nomask() +
geom_mask() +
geom_point(size=2) 
Loss on ignition
We did loss on ignition analysis both in the traditional manual method and with a modern TGA for comparison.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggplot2);
library(ggrepel);
library(tidyr);
library(reshape2) # Melt-function for reshaping data to long form
loi <- read.xlsx("data/loi_AY.xlsx", sep=";")
tga <- read.xlsx("data/tga_AY.xlsx", sep=";")
summary(loi)## sample crucible wet_weight sample_wet
## Length:52 Min. :8.079 Min. :10.13 Min. :1.538
## Class :character 1st Qu.:8.566 1st Qu.:11.00 1st Qu.:2.464
## Mode :character Median :9.006 Median :11.71 Median :2.635
## Mean :8.994 Mean :11.55 Mean :2.559
## 3rd Qu.:9.435 3rd Qu.:12.05 3rd Qu.:2.824
## Max. :9.925 Max. :12.68 Max. :3.043
## dry_weight after_550_C after_950_C context
## Min. :10.11 Min. :10.08 Min. :10.05 Length:52
## 1st Qu.:10.97 1st Qu.:10.94 1st Qu.:10.86 Class :character
## Median :11.67 Median :11.62 Median :11.53 Mode :character
## Mean :11.52 Mean :11.47 Mean :11.41
## 3rd Qu.:12.00 3rd Qu.:11.96 3rd Qu.:11.93
## Max. :12.65 Max. :12.60 Max. :12.52
## type
## Length:52
## Class :character
## Mode :character
##
##
## summary(tga)## Name Analysis;Date Batch Wet
## Length:53 Length:53 Length:53 Length:53
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## Dry Mass_550 Mass_950 context
## Length:53 Length:53 Length:53 Length:53
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## type
## Length:53
## Class :character
## Mode :character# Remove rows containing words 'kekkila' (internal standard)
tga <- filter(tga, !grepl('kekkila', Name))
# TGA character columns to numeric
tga[, c(4:7)] <- sapply(tga[, c(4:7)], as.numeric)
# Sample names as rownames
rownames(loi) <- loi$sample
rownames(tga) <- tga$Name
# Creating new columns for analysis by subtracting crucible mass;
# "dry" (sample mass after drying)
# "c550" (sample mass after 550 C)
# "c950" (sample mass after 950 C)
loi$dry <- (loi$dry_weight - loi$crucible)
loi$c550mass <- (loi$after_550_C - loi$crucible)
loi$c950mass <- (loi$after_950_C - loi$crucible)
loi$c550 <- (loi$dry - loi$c550mass)/(loi$dry) * 100
loi$c950 <- (loi$c550mass - loi$c950mass)/(loi$c550mass) * 100
# Same with TGA
tga$c550 <- (tga$Dry - tga$Mass_550)/(tga$Dry) * 100
tga$c950 <- (tga$Mass_550 - tga$Mass_950)/(tga$Mass_550) * 100Manual LOI results
# Colored boxplots in same graph
loi_long <- loi[c(8,13:14)]
loi_long <- melt(loi_long, id = "context") # Reshaping data to long form
ggplot(loi_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Traditional LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(loi_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context",
x ="Context", y = "LOI")
# boxplot.stats gives us the lower whisker or minimum, the interquartile range of the middle (the second and fourth values), the median (third value) and the upper whisker or maximum.
boxplot.stats(loi$c950)## $stats
## [1] 0.398150 1.528018 2.198513 3.162314 4.889189
##
## $n
## [1] 52
##
## $conf
## [1] 1.840428 2.556598
##
## $out
## [1] 8.041755 11.960444# The two high carbonate outliers are both AY-53 (two different measurements to make sure the first batch wasn't misleading)
# Removing duplicates from the next graph
rownames(loi)## [1] "AY-1" "AY-2" "AY-3" "AY-4" "AY-5" "AY-6" "AY-7"
## [8] "AY-8" "AY-9" "AY-10" "AY-11" "AY-12" "AY-13" "AY-14"
## [15] "AY-15" "AY-16" "AY-17" "AY-18" "AY-19" "AY-20" "AY-21"
## [22] "AY-22" "AY-23" "AY-24" "AY-25" "AY-26" "AY-27" "AY-28"
## [29] "AY-29" "AY-30" "AY-31" "AY-32" "AY-33" "AY-34" "AY-35"
## [36] "AY-36" "AY-37" "AY-38" "AY-39" "AY-40" "AY-41" "AY-42"
## [43] "AY-50" "AY-51" "AY-52" "AY-53" "AY-54" "AY-50_2" "AY-51_2"
## [50] "AY-52_2" "AY-53_2" "AY-54_2"loi <- subset(loi[1:47, ])
# Color by context
ggplot(loi,
aes(c550, c950, label = rownames(loi), colour = factor(context))) +
geom_point(size=2, aes(shape = factor(type))) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Color by type
ggplot(loi,
aes(c550, c950)) +
geom_point(size=2, aes(shape = factor(type), colour = factor(context))) +
labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
loi_long = gather(loi, key = var, value = value, c550, c950)
ggplot(loi_long, aes(x = reorder(sample, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
TGA LOI results
# Colored boxplots in same graph
tga_long <- tga[c(8,10:11)]
tga_long <- melt(tga_long, id = "context") # Reshaping data to long form
ggplot(tga_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(tga_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Color by context biplot
ggplot(tga,
aes(c550, c950, label = rownames(tga), colour = factor(context))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
tga_long = gather(tga, key = var, value = value, c550, c950)
ggplot(tga_long, aes(x = reorder(Name, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
Colorimetry
# Libraries
library(dplyr);
library(ggplot2);
library(ggrepel);
library(rgl);
library(plot3D);
library(colordistance)
color1 <- read.table("data/color_AY.txt", header = TRUE, sep = "\t", dec = ",")
rownames(color1) <- color1$Data.Name
color <- subset(color1[, 3:5])
summary(color)## L a b
## Min. :37.56 Min. : 7.97 Min. :16.81
## 1st Qu.:42.13 1st Qu.: 9.72 1st Qu.:19.76
## Median :43.64 Median :10.38 Median :21.18
## Mean :43.92 Mean :10.49 Mean :21.22
## 3rd Qu.:45.91 3rd Qu.:11.29 3rd Qu.:22.34
## Max. :48.24 Max. :13.53 Max. :25.03ggplot(color,
aes(a, b, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Color (a*b)") +
xlab("a D65") +
ylab("b D65") 
ggplot(color1,
aes(X, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("Sample number") 
ggplot(color,
aes(a, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "A versus luminance") +
ylab("L* (D65)") +
xlab("a") 
ggplot(color,
aes(b, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "B versus luminance") +
ylab("L* (D65)") +
xlab("b") 
# Colorful 3D plot with drop lines
scatter3D(color$a, color$b, color$L,
xlab = "a* (D65)", ylab = "b* (D65)", zlab="L* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$a, color$b, color$L, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
scatter3D(color$L, color$a, color$b,
xlab = "L* (D65)", ylab = "a* (D65)", zlab="b* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$L, color$a, color$b, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
Ashdod-Yam: Byzantine
Intro about the site
pXRF
From literature we can identify three rough classes of elements that could be used to analyze mudbricks from pXRF results:
Good: Rb, Sr, Y, Zr, Nb, Th, Pb, Cu, Zn, Fe
Neutral: V, Cr, Co, Ni, Ca, Ti
Bad: Na, P, Ba (+ Ca, Al)
The “bad” elements were excluded from the pXRF results from the get-go. After scaling and averaging the remaining values we ended up with two analysis datasets: “pxrf_all” that contains all the elements that have non-zero results, and “pxrf_no_na”, in which all the elements that have even one NA result are omitted.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggfortify); # PCA autoplots
library(FactoMineR); # Fast PCA graphs
library(dendextend); # Dendrograms
# Read the raw data
pxrf <- read.xlsx("data/pxrf_AYB.xlsx", sep=";")
# Remove any rows containing words 'ERROR' or 'TEST'
pxrf <- filter(pxrf, !grepl('TEST', Sample))
pxrf <- filter(pxrf, !grepl('ERROR', Sample))
# Remove unused columns
pxrf <- pxrf %>%
select(-Site) %>%
select(-'Other;name') %>%
select(-'File;#') %>%
select(-DateTime) %>%
select(-Application) %>%
select(-Method) %>%
select(-ElapsedTime) %>%
select(-'Cal;Check')
# Coercing "< LOD" -results to NA:s
pxrf[4:93] <- pxrf[4:93] %>% mutate_if(is.character,as.numeric)
# Averaging results by "Sample", "Area" and "Type" columns
pxrf_averaged <- aggregate(pxrf, by = list(pxrf$Sample, pxrf$Area, pxrf$Type), FUN = mean)
# Removing old (now empty) columns
pxrf_averaged <- pxrf_averaged %>% select(-Sample)
pxrf_averaged <- pxrf_averaged %>% select(-Area)
pxrf_averaged <- pxrf_averaged %>% select(-Type)
# Assigning sample names as new row names
rownames(pxrf_averaged) <- pxrf_averaged$Group.1
pxrf_averaged <- pxrf_averaged %>% select(-Group.1)
# Renaming the other two newly created groups back to "Type" and "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.2"] <- "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.3"] <- "Type"
# Removing error columns from the scaled analysis data set
pxrf_scaled <- pxrf_averaged %>% select(-ends_with("Err"))
# Scaling the data with standard z-score method
pxrf_scaled[3:47] <- scale(pxrf_scaled[3:47])
# Removing columns that only contain NA values
all_na <- function(x) any(!is.na(x))
pxrf_all <- pxrf_scaled %>% select_if(all_na)pxrf_averaged: All averaged data including averaged errors (Uine’s note: for later use)
colnames(pxrf_averaged)## [1] "Area" "Type" "MgO" "MgO;Err" "Al2O3" "Al2O3;Err"
## [7] "SiO2" "SiO2;Err" "P2O5" "P2O5;Err" "S" "S;Err"
## [13] "Cl" "Cl;Err" "K2O" "K2O;Err" "CaO" "CaO;Err"
## [19] "Ti" "Ti;Err" "V" "V;Err" "Cr" "Cr;Err"
## [25] "Mn" "Mn;Err" "Fe" "Fe;Err" "Co" "Co;Err"
## [31] "Ni" "Ni;Err" "Cu" "Cu;Err" "Zn" "Zn;Err"
## [37] "As" "As;Err" "Se" "Se;Err" "Rb" "Rb;Err"
## [43] "Sr" "Sr;Err" "Y" "Y;Err" "Zr" "Zr;Err"
## [49] "Nb" "Nb;Err" "Mo" "Mo;Err" "Rh" "Rh;Err"
## [55] "Pd" "Pd;Err" "Ag" "Ag;Err" "Cd" "Cd;Err"
## [61] "Sn" "Sn;Err" "Sb" "Sb;Err" "Ba" "Ba;Err"
## [67] "La" "La;Err" "Ce" "Ce;Err" "Hf" "Hf;Err"
## [73] "Ta" "Ta;Err" "W" "W;Err" "Pt" "Pt;Err"
## [79] "Au" "Au;Err" "Hg" "Hg;Err" "Tl" "Tl;Err"
## [85] "Pb" "Pb;Err" "Bi" "Bi;Err" "Th" "Th;Err"
## [91] "U" "U;Err"pxrf_all: All elements (the numbers below are the amount of NA values)
colSums(is.na(pxrf_all))## Area Type MgO Al2O3 SiO2 P2O5 S Cl K2O CaO Ti Mn Fe
## 0 0 1 0 0 0 1 0 2 0 0 0 0
## Ni Cu Zn As Rb Sr Y Zr Pb
## 0 0 0 1 1 0 2 0 3pxrf_final: Final analysis data set with only the selected elements
# Dropping the dubious elements and elements with too many NA values
pxrf_final <- pxrf_all %>%
select(-Al2O3) %>%
select(-CaO) %>%
select(-SiO2) %>%
select(-Cl) %>%
select(-K2O) %>%
select(-Ni) %>%
select(-P2O5) %>%
select(-S) %>%
select(-As) %>%
select(-Pb)
# Transforming remaining NA values to zeros for analysis
pxrf_final[is.na(pxrf_final)] <- 0
# Final analysis data set
colnames(pxrf_final)## [1] "Area" "Type" "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr"
## [11] "Y" "Zr"pXRF: K means cluster analysis
# PC autoplot with K-means clusters
set.seed(1)
autoplot(kmeans(pxrf_final[3:12], 4), data = pxrf_final, label = TRUE, label.size = 3)
pXRF: PCA
PCA with final elements:
# Elements
colnames(pxrf_final[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_1 <- prcomp(pxrf_final[3:12])
summary(pca_1)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.3569 1.4536 0.9800 0.72263 0.38973 0.17504 1.58e-16
## Proportion of Variance 0.5952 0.2264 0.1029 0.05595 0.01627 0.00328 0.00e+00
## Cumulative Proportion 0.5952 0.8216 0.9245 0.98044 0.99672 1.00000 1.00e+00# PCA plots
biplot(pca_1, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Byzantine elements")
autoplot(pca_1, data=pxrf_final, colour='Type', shape = FALSE, label = TRUE, main = "PCA Byzantine grouped by sample type")
PCA(pxrf_final[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 7 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"pXRF: HCA
# HCA dendrogram, samples color coded by type:
dend2 <-
pxrf_final %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=5)
type <- as.factor(pxrf_final[, 2])
n_type <- length(unique(type))
cols_t <- colorspace::rainbow_hcl(n_type, c = 70, l = 50)
col_type <- cols_t[type]
labels_colors(dend2) <- col_type[order.dendrogram(dend2)]
plot(dend2, main="HCA with sample types")
rect.dendrogram(dend2, 5, border = "Black")
legend("topright", legend = levels(type), fill = cols_t)
Grain size analysis
Data preparation
# Libraries
library(dplyr);
library(openxlsx);
library(ggtern)
# Read and filter data
grain <- read.xlsx("data/grain_AYB.xlsx", sep=";")
# Remove automatically created averages in order to include results from multiple runs for the same sample
grain <- grain %>%
filter(!grepl('Average', Sample)) %>%
filter(!grepl('test', Sample))
# Average data by "Sample", "Context" and "Type" columns from the original dataset
grain <- aggregate(grain, by=list(grain$Sample, grain$Context, grain$Type), FUN=mean)
# Remove the now empty columns for clarity
grain <- grain %>% select(-Sample)
grain <- grain %>% select(-Context)
grain <- grain %>% select(-Type)
# Assign sample names ("Group.1") as row names, and rename the other created columns back to "Type" and "Area" for clarity
rownames(grain) <- grain$Group.1
grain <- grain %>% select(-Group.1)
names(grain)[names(grain) == "Group.2"] <- "Context"
names(grain)[names(grain) == "Group.3"] <- "Type"
# Scaling clay-silt-sand portions to values between 0 and 1
# (Otherwise the differences in low clay percentages are too small to differentiate)
normalize <- function(x, na.rm=TRUE){(x-min(x, na.rm=TRUE))/(max(x, na.rm=TRUE)-min(x, na.rm=TRUE))}
# Apply function "normalize" to clay, silt and sand columns
grain_01 <- as.data.frame(apply(grain[4:6], 2, normalize))Ternary plots
# Ternary plots
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=0) +
geom_text(aes(label=rownames(grain_01)), size=3)
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Type)) +
labs(title="Clay silt sand") +
theme_nomask() +
geom_mask() +
geom_point(size=2) 
Loss on ignition
We did loss on ignition analysis both in the traditional manual method and with a modern TGA for comparison.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggplot2);
library(ggrepel);
library(tidyr);
library(reshape2) # Melt-function for reshaping data to long form
loi <- read.xlsx("data/loi_AYB.xlsx", sep=";")
tga <- read.xlsx("data/tga_AYB.xlsx", sep=";")
summary(loi)## sample crucible wet_weight sample_wet
## Length:7 Min. :8.382 Min. :10.92 Min. :2.257
## Class :character 1st Qu.:8.555 1st Qu.:11.39 1st Qu.:2.613
## Mode :character Median :8.666 Median :11.50 Median :2.880
## Mean :8.943 Mean :11.68 Mean :2.739
## 3rd Qu.:9.442 3rd Qu.:12.13 3rd Qu.:2.905
## Max. :9.560 Max. :12.31 Max. :3.002
## dry_weight after_550_C after_950_C context
## Min. :10.83 Min. :10.62 Min. :10.50 Length:7
## 1st Qu.:11.33 1st Qu.:11.25 1st Qu.:11.04 Class :character
## Median :11.45 Median :11.39 Median :11.21 Mode :character
## Mean :11.63 Mean :11.55 Mean :11.24
## 3rd Qu.:12.11 3rd Qu.:12.05 3rd Qu.:11.43
## Max. :12.26 Max. :12.24 Max. :12.06
## type
## Length:7
## Class :character
## Mode :character
##
##
## summary(tga)## Name Analysis;Date Batch Wet
## Length:7 Min. :44536 Length:7 Length:7
## Class :character 1st Qu.:44536 Class :character Class :character
## Mode :character Median :44536 Mode :character Mode :character
## Mean :44536
## 3rd Qu.:44536
## Max. :44536
## Dry Mass_550 Mass_950 context
## Length:7 Length:7 Length:7 Length:7
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
## type
## Length:7
## Class :character
## Mode :character
##
##
## # Remove rows containing words 'kekkila' (internal standard)
tga <- filter(tga, !grepl('kekkila', Name))
# TGA character columns to numeric
tga[, c(4:7)] <- sapply(tga[, c(4:7)], as.numeric)
# Sample names as rownames
rownames(loi) <- loi$sample
rownames(tga) <- tga$Name
# Creating new columns for analysis by subtracting crucible mass;
# "dry" (sample mass after drying)
# "c550" (sample mass after 550 C)
# "c950" (sample mass after 950 C)
loi$dry <- (loi$dry_weight - loi$crucible)
loi$c550mass <- (loi$after_550_C - loi$crucible)
loi$c950mass <- (loi$after_950_C - loi$crucible)
loi$c550 <- (loi$dry - loi$c550mass)/(loi$dry) * 100
loi$c950 <- (loi$c550mass - loi$c950mass)/(loi$c550mass) * 100
# Same with TGA
tga$c550 <- (tga$Dry - tga$Mass_550)/(tga$Dry) * 100
tga$c950 <- (tga$Mass_550 - tga$Mass_950)/(tga$Mass_550) * 100Manual LOI results
# Colored boxplots in same graph
loi_long <- loi[c(8,13:14)]
loi_long <- melt(loi_long, id = "context") # Reshaping data to long form
ggplot(loi_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context",
x ="Context", y = "LOI")
# Color by type
ggplot(loi,
aes(c550, c950)) +
geom_point(size=2, aes(shape = factor(context), colour = factor(type))) +
labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
loi_long = gather(loi, key = var, value = value, c550, c950)
ggplot(loi_long, aes(x = reorder(sample, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
TGA LOI results
# Colored boxplots in same graph
tga_long <- tga[c(8,10:11)]
tga_long <- melt(tga_long, id = "context") # Reshaping data to long form
ggplot(tga_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Color by type biplot
ggplot(tga,
aes(c550, c950, label = rownames(tga), colour = factor(type))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
tga_long = gather(tga, key = var, value = value, c550, c950)
ggplot(tga_long, aes(x = reorder(Name, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
Colorimetry
# Libraries
library(dplyr);
library(ggplot2);
library(ggrepel);
library(rgl);
library(plot3D);
library(colordistance)
color1 <- read.table("data/color_AYB.txt", header = TRUE, sep = "\t", dec = ",")
rownames(color1) <- color1$Data.Name
color <- subset(color1[, 3:5])
summary(color)## L a b
## Min. :34.56 Min. : 2.480 Min. : 6.98
## 1st Qu.:38.93 1st Qu.: 5.215 1st Qu.:12.73
## Median :45.30 Median : 5.540 Median :13.28
## Mean :46.69 Mean : 6.211 Mean :14.56
## 3rd Qu.:50.95 3rd Qu.: 6.675 3rd Qu.:16.21
## Max. :67.20 Max. :11.680 Max. :23.75ggplot(color,
aes(a, b, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Color (a*b)") +
xlab("a D65") +
ylab("b D65") 
ggplot(color1,
aes(X, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("Sample number") 
ggplot(color,
aes(a, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "A versus luminance") +
ylab("L* (D65)") +
xlab("a") 
ggplot(color,
aes(b, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "B versus luminance") +
ylab("L* (D65)") +
xlab("b") 
# Colorful 3D plot with drop lines
scatter3D(color$a, color$b, color$L,
xlab = "a* (D65)", ylab = "b* (D65)", zlab="L* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$a, color$b, color$L, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
scatter3D(color$L, color$a, color$b,
xlab = "L* (D65)", ylab = "a* (D65)", zlab="b* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$L, color$a, color$b, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
Israel
Intro about the site
pXRF
From literature we can identify three rough classes of elements that could be used to analyze mudbricks from pXRF results:
Good: Rb, Sr, Y, Zr, Nb, Th, Pb, Cu, Zn, Fe
Neutral: V, Cr, Co, Ni, Ca, Ti
Bad: Na, P, Ba (+ Ca, Al)
The “bad” elements were excluded from the pXRF results from the get-go. After scaling and averaging the remaining values we ended up with two analysis datasets: “pxrf_all” that contains all the elements that have non-zero results, and “pxrf_no_na”, in which all the elements that have even one NA result are omitted.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggfortify); # PCA autoplots
library(FactoMineR); # Fast PCA graphs
library(dendextend); # Dendrograms
# Read the raw data
pxrf <- read.xlsx("data/pxrf_israel.xlsx", sep=";")
# Remove any rows containing words 'ERROR' or 'TEST'
pxrf <- filter(pxrf, !grepl('TEST', Sample))
pxrf <- filter(pxrf, !grepl('ERROR', Sample))
# Remove unused columns
pxrf <- pxrf %>%
select(-Site) %>%
select(-'Other;name') %>%
select(-'File;#') %>%
select(-DateTime) %>%
select(-Application) %>%
select(-Method) %>%
select(-ElapsedTime) %>%
select(-'Cal;Check')
# Coercing "< LOD" -results to NA:s
pxrf[4:93] <- pxrf[4:93] %>% mutate_if(is.character,as.numeric)
# Averaging results by "Sample", "Area" and "Type" columns
pxrf_averaged <- aggregate(pxrf, by = list(pxrf$Sample, pxrf$Area, pxrf$Type), FUN = mean)
# Removing old (now empty) columns
pxrf_averaged <- pxrf_averaged %>% select(-Sample)
pxrf_averaged <- pxrf_averaged %>% select(-Area)
pxrf_averaged <- pxrf_averaged %>% select(-Type)
# Assigning sample names as new row names
rownames(pxrf_averaged) <- pxrf_averaged$Group.1
pxrf_averaged <- pxrf_averaged %>% select(-Group.1)
# Renaming the other two newly created groups back to "Type" and "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.2"] <- "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.3"] <- "Type"
# Removing error columns from the scaled analysis data set
pxrf_scaled <- pxrf_averaged %>% select(-ends_with("Err"))
# Scaling the data with standard z-score method
pxrf_scaled[3:47] <- scale(pxrf_scaled[3:47])
# Removing columns that only contain NA values
all_na <- function(x) any(!is.na(x))
pxrf_all <- pxrf_scaled %>% select_if(all_na)pxrf_averaged: All averaged data including averaged errors (Uine’s note: for later use)
colnames(pxrf_averaged)## [1] "Area" "Type" "MgO" "MgO;Err" "Al2O3" "Al2O3;Err"
## [7] "SiO2" "SiO2;Err" "P2O5" "P2O5;Err" "S" "S;Err"
## [13] "Cl" "Cl;Err" "K2O" "K2O;Err" "CaO" "CaO;Err"
## [19] "Ti" "Ti;Err" "V" "V;Err" "Cr" "Cr;Err"
## [25] "Mn" "Mn;Err" "Fe" "Fe;Err" "Co" "Co;Err"
## [31] "Ni" "Ni;Err" "Cu" "Cu;Err" "Zn" "Zn;Err"
## [37] "As" "As;Err" "Se" "Se;Err" "Rb" "Rb;Err"
## [43] "Sr" "Sr;Err" "Y" "Y;Err" "Zr" "Zr;Err"
## [49] "Nb" "Nb;Err" "Mo" "Mo;Err" "Rh" "Rh;Err"
## [55] "Pd" "Pd;Err" "Ag" "Ag;Err" "Cd" "Cd;Err"
## [61] "Sn" "Sn;Err" "Sb" "Sb;Err" "Ba" "Ba;Err"
## [67] "La" "La;Err" "Ce" "Ce;Err" "Hf" "Hf;Err"
## [73] "Ta" "Ta;Err" "W" "W;Err" "Pt" "Pt;Err"
## [79] "Au" "Au;Err" "Hg" "Hg;Err" "Tl" "Tl;Err"
## [85] "Pb" "Pb;Err" "Bi" "Bi;Err" "Th" "Th;Err"
## [91] "U" "U;Err"pxrf_all: All elements (the numbers below are the amount of NA values)
colSums(is.na(pxrf_all))## Area Type MgO Al2O3 SiO2 P2O5 S Cl K2O CaO Ti Mn Fe
## 0 0 0 0 1 0 3 0 8 0 0 0 0
## Ni Cu Zn Rb Sr Y Zr Nb
## 11 0 0 6 0 16 0 17pxrf_final: Final analysis data set with only the selected elements
# Dropping the dubious elements and elements with too many NA values
pxrf_final <- pxrf_all %>%
select(-Al2O3) %>%
select(-CaO) %>%
select(-SiO2) %>%
select(-Cl) %>%
select(-K2O) %>%
select(-Ni) %>%
select(-P2O5) %>%
select(-S) %>%
select(-Nb)
# Transforming remaining NA values to zeros for analysis
pxrf_final[is.na(pxrf_final)] <- 0
# Final analysis data set
colnames(pxrf_final)## [1] "Area" "Type" "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr"
## [11] "Y" "Zr"pxrf_MB: Final analysis data set for MUDBRICK samples only (= no floor or lime plaster)
pxrf_MB <- pxrf_final[-c(1:7), ]
rownames(pxrf_MB)## [1] "AH-4" "AH-5" "RLZ-1" "TI-1" "TI-10" "TI-2" "TI-3" "TI-4" "TI-5"
## [10] "TI-6" "TI-7" "TI-8"pXRF: K means cluster analysis
# PC autoplot with K-means clusters
set.seed(1)
autoplot(kmeans(pxrf_final[3:12], 7), data = pxrf_final, label = TRUE, label.size = 3)
pXRF: PCA
PCA with final elements:
# Elements
colnames(pxrf_final[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_1 <- prcomp(pxrf_final[3:12])
summary(pca_1)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.0850 1.3922 1.0863 0.70334 0.64347 0.44563 0.31268
## Proportion of Variance 0.4953 0.2208 0.1344 0.05636 0.04717 0.02262 0.01114
## Cumulative Proportion 0.4953 0.7161 0.8505 0.90687 0.95404 0.97666 0.98780
## PC8 PC9 PC10
## Standard deviation 0.24332 0.18520 0.11656
## Proportion of Variance 0.00674 0.00391 0.00155
## Cumulative Proportion 0.99454 0.99845 1.00000# PCA plots
biplot(pca_1, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Israel elements")
autoplot(pca_1, data=pxrf_final, colour='Area', shape = FALSE, label = TRUE, main = "PCA Israel grouped by area")
PCA(pxrf_final[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 19 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"PCA with only mudbrick samples:
# Elements
colnames(pxrf_MB[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_2 <- prcomp(pxrf_MB[3:12])
summary(pca_2)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.3606 1.6226 0.7143 0.52086 0.42466 0.34190 0.21405
## Proportion of Variance 0.5941 0.2807 0.0544 0.02892 0.01923 0.01246 0.00488
## Cumulative Proportion 0.5941 0.8748 0.9292 0.95809 0.97732 0.98978 0.99467
## PC8 PC9 PC10
## Standard deviation 0.18778 0.10943 0.05281
## Proportion of Variance 0.00376 0.00128 0.00030
## Cumulative Proportion 0.99843 0.99970 1.00000# PCA plots
biplot(pca_2, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Israel elements")
autoplot(pca_2, data=pxrf_MB, colour='Area', shape = FALSE, label = TRUE, main = "PCA Israel grouped by area")
PCA(pxrf_MB[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 12 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"pXRF: HCA
# HCA dendrogram, samples color coded by area:
dend <-
pxrf_MB %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=4)
area <- as.factor(pxrf_MB[, 1])
n_area <- length(unique(area))
cols_a <- colorspace::rainbow_hcl(n_area, c = 70, l = 50)
col_area <- cols_a[area]
labels_colors(dend) <- col_area[order.dendrogram(dend)]
plot(dend, main="HCA with sample areas")
rect.dendrogram(dend, 4, border = "Black")
legend("topright", legend = levels(area), fill = cols_a)
Grain size analysis
Data preparation
# Libraries
library(dplyr);
library(openxlsx);
library(ggtern)
# Read and filter data
grain <- read.xlsx("data/grain_israel.xlsx", sep=";")
# Remove automatically created averages in order to include results from multiple runs for the same sample
grain <- grain %>%
filter(!grepl('Average', Sample)) %>%
filter(!grepl('test', Sample))
# Average data by "Sample", "Context" and "Type" columns from the original dataset
grain <- aggregate(grain, by=list(grain$Sample, grain$Context, grain$Type), FUN=mean)
# Remove the now empty columns for clarity
grain <- grain %>% select(-Sample)
grain <- grain %>% select(-Context)
grain <- grain %>% select(-Type)
# Assign sample names ("Group.1") as row names, and rename the other created columns back to "Type" and "Area" for clarity
rownames(grain) <- grain$Group.1
grain <- grain %>% select(-Group.1)
names(grain)[names(grain) == "Group.2"] <- "Context"
names(grain)[names(grain) == "Group.3"] <- "Type"
# Scaling clay-silt-sand portions to values between 0 and 1
# (Otherwise the differences in low clay percentages are too small to differentiate)
normalize <- function(x, na.rm=TRUE){(x-min(x, na.rm=TRUE))/(max(x, na.rm=TRUE)-min(x, na.rm=TRUE))}
# Apply function "normalize" to clay, silt and sand columns
grain_01 <- as.data.frame(apply(grain[4:6], 2, normalize))Ternary plots
# Ternary plots
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=0) +
geom_text(aes(label=rownames(grain_01)), size=3)
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_nomask() +
geom_mask() +
geom_point(size=2)
Loss on ignition
We did loss on ignition analysis both in the traditional manual method and with a modern TGA for comparison.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggplot2);
library(ggrepel);
library(tidyr);
library(reshape2) # Melt-function for reshaping data to long form
loi <- read.xlsx("data/loi_israel.xlsx", sep=";")
tga <- read.xlsx("data/tga_israel.xlsx", sep=";")
summary(loi)## sample crucible wet_weight sample_wet
## Length:19 Min. :7.921 Min. :10.55 Min. :2.081
## Class :character 1st Qu.:8.557 1st Qu.:11.23 1st Qu.:2.403
## Mode :character Median :9.278 Median :11.78 Median :2.584
## Mean :9.017 Mean :11.63 Mean :2.611
## 3rd Qu.:9.452 3rd Qu.:11.92 3rd Qu.:2.828
## Max. :9.787 Max. :12.62 Max. :3.231
## dry_weight after_550_C after_950_C context
## Min. :10.52 Min. :10.42 Min. : 9.773 Length:19
## 1st Qu.:11.23 1st Qu.:11.18 1st Qu.:10.376 Class :character
## Median :11.76 Median :11.70 Median :10.971 Mode :character
## Mean :11.60 Mean :11.53 Mean :10.884
## 3rd Qu.:11.88 3rd Qu.:11.78 3rd Qu.:11.322
## Max. :12.56 Max. :12.49 Max. :12.371
## type
## Length:19
## Class :character
## Mode :character
##
##
## summary(tga)## Name Analysis;Date Batch Wet
## Length:12 Length:12 Length:12 Length:12
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## Dry Mass_550 Mass_950 context
## Length:12 Length:12 Length:12 Length:12
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## type
## Length:12
## Class :character
## Mode :character# Remove rows containing words 'kekkila' (internal standard)
tga <- filter(tga, !grepl('kekkila', Name))
# TGA character columns to numeric
tga[, c(4:7)] <- sapply(tga[, c(4:7)], as.numeric)
# Sample names as rownames
rownames(loi) <- loi$sample
rownames(tga) <- tga$Name
# Creating new columns for analysis by subtracting crucible mass;
# "dry" (sample mass after drying)
# "c550" (sample mass after 550 C)
# "c950" (sample mass after 950 C)
loi$dry <- (loi$dry_weight - loi$crucible)
loi$c550mass <- (loi$after_550_C - loi$crucible)
loi$c950mass <- (loi$after_950_C - loi$crucible)
loi$c550 <- (loi$dry - loi$c550mass)/(loi$dry) * 100
loi$c950 <- (loi$c550mass - loi$c950mass)/(loi$c550mass) * 100
# Same with TGA
tga$c550 <- (tga$Dry - tga$Mass_550)/(tga$Dry) * 100
tga$c950 <- (tga$Mass_550 - tga$Mass_950)/(tga$Mass_550) * 100Manual LOI results
NOTE! Remember to remove Ah-Halom floor and plaster samples from boxplots!
# Colored boxplots in same graph
loi_long <- loi[c(8,13:14)]
loi_long <- melt(loi_long, id = "context") # Reshaping data to long form
ggplot(loi_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Traditional LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(loi_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context",
x ="Context", y = "LOI")
# Color by context
ggplot(loi,
aes(c550, c950, label = rownames(loi), colour = factor(context))) +
geom_point(size=2, aes(shape = factor(type))) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Color by type
ggplot(loi,
aes(c550, c950)) +
geom_point(size=2, aes(shape = factor(context), colour = factor(type))) +
labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
loi_long = gather(loi, key = var, value = value, c550, c950)
ggplot(loi_long, aes(x = reorder(sample, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
TGA LOI results
# Colored boxplots in same graph
tga_long <- tga[c(8,10:11)]
tga_long <- melt(tga_long, id = "context") # Reshaping data to long form
ggplot(tga_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(tga_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Color by context biplot
ggplot(tga,
aes(c550, c950, label = rownames(tga), colour = factor(context))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
tga_long = gather(tga, key = var, value = value, c550, c950)
ggplot(tga_long, aes(x = reorder(Name, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
Colorimetry
# Libraries
library(dplyr);
library(ggplot2);
library(ggrepel);
library(rgl);
library(plot3D);
library(colordistance)
color1 <- read.table("data/color_israel.txt", header = TRUE, sep = "\t", dec = ",")
rownames(color1) <- color1$Data.Name
color <- subset(color1[, 3:5])
summary(color)## L a b
## Min. :43.75 Min. : 3.160 Min. :13.85
## 1st Qu.:56.08 1st Qu.: 3.770 1st Qu.:14.71
## Median :64.30 Median : 4.350 Median :15.85
## Mean :63.19 Mean : 6.649 Mean :17.27
## 3rd Qu.:71.51 3rd Qu.: 6.280 3rd Qu.:17.70
## Max. :75.34 Max. :19.530 Max. :25.84ggplot(color,
aes(a, b, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Color (a*b)") +
xlab("a D65") +
ylab("b D65") 
ggplot(color1,
aes(X, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("Sample number") 
ggplot(color,
aes(a, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "A versus luminance") +
ylab("L* (D65)") +
xlab("a") 
ggplot(color,
aes(b, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "B versus luminance") +
ylab("L* (D65)") +
xlab("b") 
# Colorful 3D plot with drop lines
scatter3D(color$a, color$b, color$L,
xlab = "a* (D65)", ylab = "b* (D65)", zlab="L* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$a, color$b, color$L, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
scatter3D(color$L, color$a, color$b,
xlab = "L* (D65)", ylab = "a* (D65)", zlab="b* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$L, color$a, color$b, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
Palaepaphos
Intro about the site
pXRF
From literature we can identify three rough classes of elements that could be used to analyze mudbricks from pXRF results:
Good: Rb, Sr, Y, Zr, Nb, Th, Pb, Cu, Zn, Fe
Neutral: V, Cr, Co, Ni, Ca, Ti
Bad: Na, P, Ba (+ Ca, Al)
The “bad” elements were excluded from the pXRF results from the get-go. After scaling and averaging the remaining values we ended up with two analysis datasets: “pxrf_all” that contains all the elements that have non-zero results, and “pxrf_no_na”, in which all the elements that have even one NA result are omitted.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggfortify); # PCA autoplots
library(FactoMineR); # Fast PCA graphs
library(dendextend); # Dendrograms
# Read the raw data
pxrf <- read.xlsx("data/pxrf_PP.xlsx", sep=";")
# Remove any rows containing words 'ERROR' or 'TEST'
pxrf <- filter(pxrf, !grepl('TEST', Sample))
pxrf <- filter(pxrf, !grepl('ERROR', Sample))
# Remove unused columns
pxrf <- pxrf %>%
select(-Site) %>%
select(-'Other;name') %>%
select(-'File;#') %>%
select(-DateTime) %>%
select(-Application) %>%
select(-Method) %>%
select(-ElapsedTime) %>%
select(-'Cal;Check')
# Coercing "< LOD" -results to NA:s
pxrf[4:93] <- pxrf[4:93] %>% mutate_if(is.character,as.numeric)
# Averaging results by "Sample", "Area" and "Type" columns
pxrf_averaged <- aggregate(pxrf, by = list(pxrf$Sample, pxrf$Area, pxrf$Type), FUN = mean)
# Removing old (now empty) columns
pxrf_averaged <- pxrf_averaged %>% select(-Sample)
pxrf_averaged <- pxrf_averaged %>% select(-Area)
pxrf_averaged <- pxrf_averaged %>% select(-Type)
# Assigning sample names as new row names
rownames(pxrf_averaged) <- pxrf_averaged$Group.1
pxrf_averaged <- pxrf_averaged %>% select(-Group.1)
# Renaming the other two newly created groups back to "Type" and "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.2"] <- "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.3"] <- "Type"
# Removing error columns from the scaled analysis data set
pxrf_scaled <- pxrf_averaged %>% select(-ends_with("Err"))
# Scaling the data with standard z-score method
pxrf_scaled[3:47] <- scale(pxrf_scaled[3:47])
# Removing columns that only contain NA values
all_na <- function(x) any(!is.na(x))
pxrf_all <- pxrf_scaled %>% select_if(all_na)pxrf_averaged: All averaged data including averaged errors (Uine’s note: for later use)
colnames(pxrf_averaged)## [1] "Area" "Type" "MgO" "MgO;Err" "Al2O3" "Al2O3;Err"
## [7] "SiO2" "SiO2;Err" "P2O5" "P2O5;Err" "S" "S;Err"
## [13] "Cl" "Cl;Err" "K2O" "K2O;Err" "CaO" "CaO;Err"
## [19] "Ti" "Ti;Err" "V" "V;Err" "Cr" "Cr;Err"
## [25] "Mn" "Mn;Err" "Fe" "Fe;Err" "Co" "Co;Err"
## [31] "Ni" "Ni;Err" "Cu" "Cu;Err" "Zn" "Zn;Err"
## [37] "As" "As;Err" "Se" "Se;Err" "Rb" "Rb;Err"
## [43] "Sr" "Sr;Err" "Y" "Y;Err" "Zr" "Zr;Err"
## [49] "Nb" "Nb;Err" "Mo" "Mo;Err" "Rh" "Rh;Err"
## [55] "Pd" "Pd;Err" "Ag" "Ag;Err" "Cd" "Cd;Err"
## [61] "Sn" "Sn;Err" "Sb" "Sb;Err" "Ba" "Ba;Err"
## [67] "La" "La;Err" "Ce" "Ce;Err" "Hf" "Hf;Err"
## [73] "Ta" "Ta;Err" "W" "W;Err" "Pt" "Pt;Err"
## [79] "Au" "Au;Err" "Hg" "Hg;Err" "Tl" "Tl;Err"
## [85] "Pb" "Pb;Err" "Bi" "Bi;Err" "Th" "Th;Err"
## [91] "U" "U;Err"pxrf_all: All elements (the numbers below are the amount of NA values)
colSums(is.na(pxrf_all))## Area Type MgO Al2O3 SiO2 P2O5 S Cl K2O CaO Ti V Mn
## 0 0 34 34 34 35 35 34 36 34 0 41 0
## Fe Ni Cu Zn As Rb Sr Y Zr Rh Ag
## 0 0 0 34 48 35 34 36 0 51 49pxrf_final: Final analysis data set with only the selected elements, all samples (soil and published included)
# Dropping the dubious elements and elements with too many NA values
pxrf_final <- pxrf_all %>%
select(-Al2O3) %>%
select(-CaO) %>%
select(-SiO2) %>%
select(-Cl) %>%
select(-K2O) %>%
select(-Ni) %>%
select(-P2O5) %>%
select(-S) %>%
select(-V) %>%
select(-As) %>%
select(-Rh) %>%
select(-Ag)
# Transforming remaining NA values to zeros for analysis
pxrf_final[is.na(pxrf_final)] <- 0
# Final analysis data set
colnames(pxrf_final)## [1] "Area" "Type" "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr"
## [11] "Y" "Zr"pxrf_MB: Final analysis data set for NEW MUDBRICK samples only (= no soil or published samples)
pxrf_MB <- pxrf_final[c(1:11), ]
rownames(pxrf_MB)## [1] "PP-10" "PP-11" "PP-12" "PP-13" "PP-9" "PP-19" "PP-14" "PP-15" "PP-16"
## [10] "PP-17" "PP-18"pxrf_soil: Final analysis data set for NEW samples only (= no published samples, yes soil)
pxrf_soil <- pxrf_final[-c(12:45), ]
rownames(pxrf_soil)## [1] "PP-10" "PP-11" "PP-12" "PP-13" "PP-9" "PP-19" "PP-14" "PP-15" "PP-16"
## [10] "PP-17" "PP-18" "PP-1" "PP-2" "PP-3" "PP-6" "PP-4" "PP-7" "PP-5"
## [19] "PP-8"pxrf_MB2: Final analysis data set for ALL MUDBRICK samples (= no soil)
pxrf_MB2 <- pxrf_final[c(1:45), ]
rownames(pxrf_MB2)## [1] "PP-10" "PP-11" "PP-12" "PP-13" "PP-9" "PP-19" "PP-14" "PP-15" "PP-16"
## [10] "PP-17" "PP-18" "1" "10" "11" "12" "13" "14" "15"
## [19] "16" "17" "18" "19" "2" "20" "21" "22" "23"
## [28] "24" "25" "26" "27" "28" "29" "3" "30" "31"
## [37] "32" "33" "34" "4" "5" "6" "7" "8" "9"pXRF: K means cluster analysis
Only the “new” mudbrick samples
# PC autoplot with K-means clusters
set.seed(1)
autoplot(kmeans(pxrf_MB[3:12], 3), data = pxrf_MB, label = TRUE, label.size = 3)
pXRF: PCA
PCA with only new mudbrick samples:
# Elements
colnames(pxrf_MB[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_1 <- prcomp(pxrf_MB[3:12])
summary(pca_1)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.5967 1.0886 0.99328 0.85704 0.40479 0.33892 0.20666
## Proportion of Variance 0.6735 0.1184 0.09854 0.07336 0.01637 0.01147 0.00427
## Cumulative Proportion 0.6735 0.7918 0.89037 0.96373 0.98010 0.99157 0.99584
## PC8 PC9 PC10
## Standard deviation 0.17917 0.09521 0.02255
## Proportion of Variance 0.00321 0.00091 0.00005
## Cumulative Proportion 0.99904 0.99995 1.00000# PCA plots
biplot(pca_1, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Palaepaphos elements")
autoplot(pca_1, data=pxrf_MB, colour='Area', shape = FALSE, label = TRUE, main = "PCA Palaepaphos grouped by area")
PCA(pxrf_MB[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 11 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"PCA with soil samples:
# Elements
colnames(pxrf_soil[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_2 <- prcomp(pxrf_soil[3:12])
summary(pca_2)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.0533 1.6050 1.0187 0.82293 0.69376 0.39645 0.33871
## Proportion of Variance 0.4498 0.2748 0.1107 0.07225 0.05135 0.01677 0.01224
## Cumulative Proportion 0.4498 0.7247 0.8354 0.90763 0.95898 0.97575 0.98799
## PC8 PC9 PC10
## Standard deviation 0.24806 0.21529 0.0685
## Proportion of Variance 0.00657 0.00495 0.0005
## Cumulative Proportion 0.99455 0.99950 1.0000# PCA plots
biplot(pca_2, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Palaepaphos elements")
autoplot(pca_2, data=pxrf_soil, colour='Area', shape = FALSE, label = TRUE, main = "PCA Palaepaphos grouped by area")
autoplot(pca_2, data=pxrf_soil, colour='Type', shape = FALSE, label = TRUE, main = "PCA Palaepaphos grouped by sample type")
PCA(pxrf_soil[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 19 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"PCA with all the mudbrick samples
Including the previously published ones, no soil
# Elements
colnames(pxrf_MB2[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_3 <- prcomp(pxrf_MB2[3:12])
summary(pca_3)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 1.8473 1.1712 0.76106 0.60306 0.44231 0.40557 0.30364
## Proportion of Variance 0.5429 0.2182 0.09215 0.05786 0.03112 0.02617 0.01467
## Cumulative Proportion 0.5429 0.7611 0.85326 0.91112 0.94224 0.96841 0.98307
## PC8 PC9 PC10
## Standard deviation 0.22882 0.19928 0.11965
## Proportion of Variance 0.00833 0.00632 0.00228
## Cumulative Proportion 0.99140 0.99772 1.00000# PCA plots
biplot(pca_3, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Palaepaphos elements")
autoplot(pca_3, data=pxrf_MB2, colour='Area', shape = FALSE, label = TRUE, main = "PCA Palaepaphos grouped by area")
PCA(pxrf_MB2[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 45 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"pXRF: HCA
HCA including only new samples:
# HCA dendrogram, samples color coded by area:
dend <-
pxrf_MB %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=6)
area <- as.factor(pxrf_MB[, 1])
n_area <- length(unique(area))
cols_a <- colorspace::rainbow_hcl(n_area, c = 70, l = 50)
col_area <- cols_a[area]
labels_colors(dend) <- col_area[order.dendrogram(dend)]
plot(dend, main="HCA with sample areas")
rect.dendrogram(dend, 4, border = "Black")
legend("topright", legend = levels(area), fill = cols_a)
# HCA dendrogram, samples color coded by type:
dend2 <-
pxrf_soil %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=6)
type <- as.factor(pxrf_soil[, 2])
n_type <- length(unique(type))
cols_t <- colorspace::rainbow_hcl(n_type, c = 70, l = 50)
col_type <- cols_t[type]
labels_colors(dend2) <- col_type[order.dendrogram(dend2)]
plot(dend2, main="HCA with new samples by type")
rect.dendrogram(dend2, 6, border = "Black")
legend("topright", legend = levels(type), fill = cols_t)
HCA including all mudbrick samples :
# HCA dendrogram, samples color coded by area:
dend3 <-
pxrf_MB2 %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=4)
area <- as.factor(pxrf_MB2[, 1])
n_area <- length(unique(area))
cols_a <- colorspace::rainbow_hcl(n_area, c = 70, l = 50)
col_area <- cols_a[area]
labels_colors(dend3) <- col_area[order.dendrogram(dend3)]
plot(dend3, main="HCA with sample areas")
rect.dendrogram(dend3, 4, border = "Black")
legend("topright", legend = levels(area), fill = cols_a)
Grain size analysis
Data preparation
# Libraries
library(dplyr);
library(openxlsx);
library(ggtern)
# Read and filter data
grain <- read.xlsx("data/grain_PP.xlsx", sep=";")
# Remove automatically created averages in order to include results from multiple runs for the same sample
grain <- grain %>%
filter(!grepl('Average', Sample)) %>%
filter(!grepl('test', Sample))
# Average data by "Sample", "Context" and "Type" columns from the original dataset
grain <- aggregate(grain, by=list(grain$Sample, grain$Context, grain$Type), FUN=mean)
# Remove the now empty columns for clarity
grain <- grain %>% select(-Sample)
grain <- grain %>% select(-Context)
grain <- grain %>% select(-Type)
# Assign sample names ("Group.1") as row names, and rename the other created columns back to "Type" and "Area" for clarity
rownames(grain) <- grain$Group.1
grain <- grain %>% select(-Group.1)
names(grain)[names(grain) == "Group.2"] <- "Context"
names(grain)[names(grain) == "Group.3"] <- "Type"
# Scaling clay-silt-sand portions to values between 0 and 1
# (Otherwise the differences in low clay percentages are too small to differentiate)
normalize <- function(x, na.rm=TRUE){(x-min(x, na.rm=TRUE))/(max(x, na.rm=TRUE)-min(x, na.rm=TRUE))}
# Apply function "normalize" to clay, silt and sand columns
grain_01 <- as.data.frame(apply(grain[4:6], 2, normalize))Ternary plots
# Creating subsets for only mudbricks
MB_grain <- grain_01[c(1:11),]
MB_context <- grain[c(1:11),]
# Ternary plots
ggtern(data=MB_grain, aes(x=Clay, y=Sand, z=Silt, color=MB_context$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=0) +
geom_text(aes(label=rownames(MB_grain)), size=3)
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_nomask() +
geom_mask() +
geom_point(size=2) 
Loss on ignition
We did loss on ignition analysis both in the traditional manual method and with a modern TGA for comparison.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggplot2);
library(ggrepel);
library(tidyr);
library(reshape2) # Melt-function for reshaping data to long form
loi <- read.xlsx("data/loi_PP.xlsx", sep=";")
tga <- read.xlsx("data/tga_PP.xlsx", sep=";")
summary(loi)## sample crucible wet;weight sample;wet
## Length:27 Min. :8.100 Min. : 9.637 Min. :1.537
## Class :character 1st Qu.:8.639 1st Qu.:10.859 1st Qu.:2.193
## Mode :character Median :8.997 Median :11.184 Median :2.371
## Mean :8.988 Mean :11.274 Mean :2.286
## 3rd Qu.:9.454 3rd Qu.:11.844 3rd Qu.:2.453
## Max. :9.560 Max. :12.301 Max. :2.785
## dry_weight after_550_C after_950_C context
## Min. : 9.595 Min. : 9.484 Min. : 9.23 Length:27
## 1st Qu.:10.813 1st Qu.:10.721 1st Qu.:10.27 Class :character
## Median :11.132 Median :11.019 Median :10.59 Mode :character
## Mean :11.216 Mean :11.113 Mean :10.67
## 3rd Qu.:11.777 3rd Qu.:11.681 3rd Qu.:11.18
## Max. :12.218 Max. :12.076 Max. :11.56
## type
## Length:27
## Class :character
## Mode :character
##
##
## summary(tga)## Name Analysis;Date Batch Wet
## Length:30 Length:30 Length:30 Length:30
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## Dry Mass_550 Mass_950 context
## Length:30 Length:30 Length:30 Length:30
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## type
## Length:30
## Class :character
## Mode :character# Remove rows containing words 'kekkila' (internal standard)
tga <- filter(tga, !grepl('kekkila', Name))
# TGA character columns to numeric
tga[, c(4:7)] <- sapply(tga[, c(4:7)], as.numeric)
# Sample names as rownames
rownames(loi) <- loi$sample
rownames(tga) <- tga$Name
# Creating new columns for analysis by subtracting crucible mass;
# "dry" (sample mass after drying)
# "c550" (sample mass after 550 C)
# "c950" (sample mass after 950 C)
loi$dry <- (loi$dry_weight - loi$crucible)
loi$c550mass <- (loi$after_550_C - loi$crucible)
loi$c950mass <- (loi$after_950_C - loi$crucible)
loi$c550 <- (loi$dry - loi$c550mass)/(loi$dry) * 100
loi$c950 <- (loi$c550mass - loi$c950mass)/(loi$c550mass) * 100
# Same with TGA
tga$c550 <- (tga$Dry - tga$Mass_550)/(tga$Dry) * 100
tga$c950 <- (tga$Mass_550 - tga$Mass_950)/(tga$Mass_550) * 100Manual LOI results
# Removing duplicates and creating MB only subset
loi2 <- subset(loi[1:19, ])
loi_MB <- subset(loi2[9:19, ])
rownames(loi_MB)## [1] "PP-9" "PP-10" "PP-11" "PP-12" "PP-13" "PP-14" "PP-15" "PP-16" "PP-17"
## [10] "PP-18" "PP-19"# Colored boxplots in same graph
loi_MB_long <- loi_MB[c(8,13:14)]
loi_MB_long <- melt(loi_MB_long, id = "context") # Reshaping data to long form
ggplot(loi_MB_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Traditional LOI by temperature (MB only)",
x ="Temperature", y = "LOI")
ggplot(loi_MB_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context (MB only)",
x ="Context", y = "LOI")
# Colored boxplots in same graph: Soil samples included
loi_long <- loi[c(8,13:14)]
loi_long <- melt(loi_long, id = "context") # Reshaping data to long form
ggplot(loi_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Traditional LOI by temperature (soil included)",
x ="Temperature", y = "LOI")
ggplot(loi_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context (soil included)",
x ="Context", y = "LOI")
# MB only biplot
ggplot(loi_MB,
aes(c550, c950, label = rownames(loi_MB), colour = factor(context))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "MB only organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Color by context biplot
ggplot(loi,
aes(c550, c950, label = rownames(loi), colour = factor(context))) +
geom_point(size=2, aes(shape = factor(type))) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Color by type biplot
ggplot(loi,
aes(c550, c950)) +
geom_point(size=2, aes(colour = factor(type))) +
labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
loi_long = gather(loi_MB, key = var, value = value, c550, c950)
ggplot(loi_long, aes(x = reorder(sample, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
TGA LOI results
# Creating MB only subset
tga_MB1 <- subset(tga[9:22, ])
# Colored boxplots in same graph
tga_MB_long <- tga_MB1[c(8,10:11)]
tga_MB_long <- melt(tga_MB_long, id = "context") # Reshaping data to long form
ggplot(tga_MB_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(tga_MB_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Colored boxplots in same graph: soil samples included
tga_long <- tga[c(8,10:11)]
tga_long <- melt(tga_long, id = "context") # Reshaping data to long form
ggplot(tga_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(tga_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Removing duplicates and creating MB only subset
tga_MB <- subset(tga[9:20, ])
tga_MB <- subset(tga_MB[-4, ])
rownames(tga_MB)## [1] "PP-9" "PP-10" "PP-11" "PP-12" "PP-13" "PP-14" "PP-15" "PP-16" "PP-17"
## [10] "PP-18" "PP-19"# Color by context biplot
ggplot(tga_MB,
aes(c550, c950, label = rownames(tga_MB), colour = factor(context))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
tga_long = gather(tga_MB, key = var, value = value, c550, c950)
ggplot(tga_long, aes(x = reorder(Name, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
Colorimetry
# Libraries
library(dplyr);
library(ggplot2);
library(ggrepel);
library(rgl);
library(plot3D);
library(colordistance)
color1 <- read.table("data/color_PP.txt", header = TRUE, sep = "\t", dec = ",")
rownames(color1) <- color1$Data.Name
color <- subset(color1[, 3:5])
summary(color)## L a b
## Min. :62.38 Min. : 7.27 Min. :19.90
## 1st Qu.:63.47 1st Qu.: 8.50 1st Qu.:21.36
## Median :64.19 Median : 9.34 Median :21.72
## Mean :65.50 Mean : 9.29 Mean :21.92
## 3rd Qu.:65.78 3rd Qu.:10.56 3rd Qu.:22.96
## Max. :72.00 Max. :10.90 Max. :23.90ggplot(color,
aes(a, b, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Color") +
xlab("a D65") +
ylab("b D65") 
ggplot(color1,
aes(X, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("Sample number") 
ggplot(color,
aes(a, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("a") 
ggplot(color,
aes(b, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("b") 
# Colorful 3D plot with drop lines
scatter3D(color$a, color$b, color$L,
xlab = "a* (D65)", ylab = "b* (D65)", zlab="L* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$a, color$b, color$L, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
scatter3D(color$L, color$a, color$b,
xlab = "L* (D65)", ylab = "a* (D65)", zlab="b* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$L, color$a, color$b, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
Artaxata
Intro about the site
pXRF
From literature we can identify three rough classes of elements that could be used to analyze mudbricks from pXRF results:
Good: Rb, Sr, Y, Zr, Nb, Th, Pb, Cu, Zn, Fe
Neutral: V, Cr, Co, Ni, Ca, Ti
Bad: Na, P, Ba (+ Ca, Al)
The “bad” elements were excluded from the pXRF results from the get-go. After scaling and averaging the remaining values we ended up with two analysis datasets: “pxrf_all” that contains all the elements that have non-zero results, and “pxrf_no_na”, in which all the elements that have even one NA result are omitted.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggfortify); # PCA autoplots
library(FactoMineR); # Fast PCA graphs
library(dendextend); # Dendrograms
# Read the raw data
pxrf <- read.xlsx("data/pxrf_AA.xlsx", sep=";")
# Remove any rows containing words 'ERROR' or 'TEST'
pxrf <- filter(pxrf, !grepl('TEST', Sample))
pxrf <- filter(pxrf, !grepl('ERROR', Sample))
# Remove unused columns
pxrf <- pxrf %>%
select(-Site) %>%
select(-'Other;name') %>%
select(-'File;#') %>%
select(-DateTime) %>%
select(-Application) %>%
select(-Method) %>%
select(-ElapsedTime) %>%
select(-'Cal;Check')
# Coercing "< LOD" -results to NA:s
pxrf[4:93] <- pxrf[4:93] %>% mutate_if(is.character,as.numeric)
# Averaging results by "Sample", "Area" and "Type" columns
pxrf_averaged <- aggregate(pxrf, by = list(pxrf$Sample, pxrf$Area, pxrf$Type), FUN = mean)
# Removing old (now empty) columns
pxrf_averaged <- pxrf_averaged %>% select(-Sample)
pxrf_averaged <- pxrf_averaged %>% select(-Area)
pxrf_averaged <- pxrf_averaged %>% select(-Type)
# Assigning sample names as new row names
rownames(pxrf_averaged) <- pxrf_averaged$Group.1
pxrf_averaged <- pxrf_averaged %>% select(-Group.1)
# Renaming the other two newly created groups back to "Type" and "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.2"] <- "Area"
names(pxrf_averaged)[names(pxrf_averaged) == "Group.3"] <- "Type"
# Removing error columns from the scaled analysis data set
pxrf_scaled <- pxrf_averaged %>% select(-ends_with("Err"))
# Scaling the data with standard z-score method
pxrf_scaled[3:47] <- scale(pxrf_scaled[3:47])
# Removing columns that only contain NA values
all_na <- function(x) any(!is.na(x))
pxrf_all <- pxrf_scaled %>% select_if(all_na)pxrf_averaged: All averaged data including averaged errors (Uine’s note: for later use)
colnames(pxrf_averaged)## [1] "Area" "Type" "MgO" "MgO;Err" "Al2O3" "Al2O3;Err"
## [7] "SiO2" "SiO2;Err" "P2O5" "P2O5;Err" "S" "S;Err"
## [13] "Cl" "Cl;Err" "K2O" "K2O;Err" "CaO" "CaO;Err"
## [19] "Ti" "Ti;Err" "V" "V;Err" "Cr" "Cr;Err"
## [25] "Mn" "Mn;Err" "Fe" "Fe;Err" "Co" "Co;Err"
## [31] "Ni" "Ni;Err" "Cu" "Cu;Err" "Zn" "Zn;Err"
## [37] "As" "As;Err" "Se" "Se;Err" "Rb" "Rb;Err"
## [43] "Sr" "Sr;Err" "Y" "Y;Err" "Zr" "Zr;Err"
## [49] "Nb" "Nb;Err" "Mo" "Mo;Err" "Rh" "Rh;Err"
## [55] "Pd" "Pd;Err" "Ag" "Ag;Err" "Cd" "Cd;Err"
## [61] "Sn" "Sn;Err" "Sb" "Sb;Err" "Ba" "Ba;Err"
## [67] "La" "La;Err" "Ce" "Ce;Err" "Hf" "Hf;Err"
## [73] "Ta" "Ta;Err" "W" "W;Err" "Pt" "Pt;Err"
## [79] "Au" "Au;Err" "Hg" "Hg;Err" "Tl" "Tl;Err"
## [85] "Pb" "Pb;Err" "Bi" "Bi;Err" "Th" "Th;Err"
## [91] "U" "U;Err"pxrf_all: All elements (the numbers below are the amount of NA values)
colSums(is.na(pxrf_all))## Area Type MgO Al2O3 SiO2 P2O5 S Cl K2O CaO Ti V Cr
## 0 0 0 0 0 1 0 0 0 0 0 5 8
## Mn Fe Ni Cu Zn As Rb Sr Y Zr Nb Ba
## 0 0 0 0 0 1 0 0 0 0 4 3pxrf_final: Final analysis data set with only the selected elements
# Dropping the dubious elements and elements with too many NA values
pxrf_final <- pxrf_all %>%
select(-Al2O3) %>%
select(-CaO) %>%
select(-SiO2) %>%
select(-Cl) %>%
select(-K2O) %>%
select(-Ni) %>%
select(-P2O5) %>%
select(-S) %>%
select(-V) %>%
select(-As) %>%
select(-Nb) %>%
select(-Cr) %>%
select(-Ba)
# Transforming remaining NA values to zeros for analysis
pxrf_final[is.na(pxrf_final)] <- 0
# Final analysis data set
colnames(pxrf_final)## [1] "Area" "Type" "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr"
## [11] "Y" "Zr"pXRF: K means cluster analysis
# PC autoplot with K-means clusters
set.seed(1)
autoplot(kmeans(pxrf_final[3:12], 4), data = pxrf_final, label = TRUE, label.size = 3)
pXRF: PCA
PCA with final elements:
# Elements
colnames(pxrf_final[3:12])## [1] "MgO" "Ti" "Mn" "Fe" "Cu" "Zn" "Rb" "Sr" "Y" "Zr"# PCA analysis
pca_1 <- prcomp(pxrf_final[3:12])
summary(pca_1)## Importance of components:
## PC1 PC2 PC3 PC4 PC5 PC6 PC7
## Standard deviation 2.6820 1.1972 0.84041 0.53728 0.45231 0.32524 0.2122
## Proportion of Variance 0.7193 0.1433 0.07063 0.02887 0.02046 0.01058 0.0045
## Cumulative Proportion 0.7193 0.8627 0.93328 0.96215 0.98261 0.99318 0.9977
## PC8 PC9 PC10
## Standard deviation 0.14003 0.05947 0
## Proportion of Variance 0.00196 0.00035 0
## Cumulative Proportion 0.99965 1.00000 1# PCA plots
biplot(pca_1, choices = 1:2, cex = c(1, 1.2), col = c("grey80", "deeppink2"), main = "PCA Artaxata elements")
autoplot(pca_1, data=pxrf_final, colour='Area', shape = FALSE, label = TRUE, main = "PCA Artaxata grouped by area")
PCA(pxrf_final[3:12])
## **Results for the Principal Component Analysis (PCA)**
## The analysis was performed on 10 individuals, described by 10 variables
## *The results are available in the following objects:
##
## name description
## 1 "$eig" "eigenvalues"
## 2 "$var" "results for the variables"
## 3 "$var$coord" "coord. for the variables"
## 4 "$var$cor" "correlations variables - dimensions"
## 5 "$var$cos2" "cos2 for the variables"
## 6 "$var$contrib" "contributions of the variables"
## 7 "$ind" "results for the individuals"
## 8 "$ind$coord" "coord. for the individuals"
## 9 "$ind$cos2" "cos2 for the individuals"
## 10 "$ind$contrib" "contributions of the individuals"
## 11 "$call" "summary statistics"
## 12 "$call$centre" "mean of the variables"
## 13 "$call$ecart.type" "standard error of the variables"
## 14 "$call$row.w" "weights for the individuals"
## 15 "$call$col.w" "weights for the variables"pXRF: HCA
# HCA dendrogram, samples color coded by area:
dend <-
pxrf_final %>% # data
dist %>% # calculate a distance matrix
hclust(method = "ward.D2") %>% # hierarchical clustering
as.dendrogram %>% # turn the object into a dendrogram
color_branches(k=5)
area <- as.factor(pxrf_final[, 1])
n_area <- length(unique(area))
cols_a <- colorspace::rainbow_hcl(n_area, c = 70, l = 50)
col_area <- cols_a[area]
labels_colors(dend) <- col_area[order.dendrogram(dend)]
plot(dend, main="HCA with sample areas")
rect.dendrogram(dend, 5, border = "Black")
legend("topright", legend = levels(area), fill = cols_a)
Grain size analysis
Data preparation
# Libraries
library(dplyr);
library(openxlsx);
library(ggtern)
# Read and filter data
grain <- read.xlsx("data/grain_AA.xlsx", sep=";")
# Remove automatically created averages in order to include results from multiple runs for the same sample
grain <- grain %>%
filter(!grepl('Average', Sample)) %>%
filter(!grepl('test', Sample))
# Average data by "Sample", "Context" and "Type" columns from the original dataset
grain <- aggregate(grain, by=list(grain$Sample, grain$Context, grain$Type), FUN=mean)
# Remove the now empty columns for clarity
grain <- grain %>% select(-Sample)
grain <- grain %>% select(-Context)
grain <- grain %>% select(-Type)
# Assign sample names ("Group.1") as row names, and rename the other created columns back to "Type" and "Area" for clarity
rownames(grain) <- grain$Group.1
grain <- grain %>% select(-Group.1)
names(grain)[names(grain) == "Group.2"] <- "Context"
names(grain)[names(grain) == "Group.3"] <- "Type"
# Scaling clay-silt-sand portions to values between 0 and 1
# (Otherwise the differences in low clay percentages are too small to differentiate)
normalize <- function(x, na.rm=TRUE){(x-min(x, na.rm=TRUE))/(max(x, na.rm=TRUE)-min(x, na.rm=TRUE))}
# Apply function "normalize" to clay, silt and sand columns
grain_01 <- as.data.frame(apply(grain[4:6], 2, normalize))Ternary plots
# Ternary plots
ggtern(data=grain_01, aes(x=Clay, y=Sand, z=Silt, color=grain$Context)) +
labs(title="Clay silt sand") +
theme_rgbw() +
theme_nomask() +
geom_point(size=0) +
geom_text(aes(label=rownames(grain_01)), size=3)
Loss on ignition
We did loss on ignition analysis both in the traditional manual method and with a modern TGA for comparison.
Data preparation
# Libraries
library(openxlsx);
library(dplyr);
library(ggplot2);
library(ggrepel);
library(tidyr);
library(reshape2) # Melt-function for reshaping data to long form
loi <- read.xlsx("data/loi_AA.xlsx", sep=";")
tga <- read.xlsx("data/tga_AA.xlsx", sep=";")
summary(loi)## sample crucible wet_weight sample_wet
## Length:10 Min. :8.066 Min. :10.70 Min. :1.923
## Class :character 1st Qu.:8.461 1st Qu.:10.93 1st Qu.:2.326
## Mode :character Median :8.995 Median :11.48 Median :2.558
## Mean :8.892 Mean :11.39 Mean :2.496
## 3rd Qu.:9.174 3rd Qu.:11.75 3rd Qu.:2.701
## Max. :9.890 Max. :12.18 Max. :2.842
## dry_weight after_550_C after_950_C context
## Min. :10.58 Min. :10.47 Min. :10.31 Length:10
## 1st Qu.:10.85 1st Qu.:10.78 1st Qu.:10.60 Class :character
## Median :11.43 Median :11.33 Median :11.16 Mode :character
## Mean :11.32 Mean :11.22 Mean :11.07
## 3rd Qu.:11.69 3rd Qu.:11.58 3rd Qu.:11.42
## Max. :12.13 Max. :12.06 Max. :11.94
## type
## Length:10
## Class :character
## Mode :character
##
##
## summary(tga)## Name Analysis;Date Batch Wet
## Length:12 Length:12 Length:12 Length:12
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## Dry Mass_550 Mass_950 context
## Length:12 Length:12 Length:12 Length:12
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
## type
## Length:12
## Class :character
## Mode :character# Remove rows containing words 'kekkila' (internal standard)
tga <- filter(tga, !grepl('kekkila', Name))
# TGA character columns to numeric
tga[, c(4:7)] <- sapply(tga[, c(4:7)], as.numeric)
# Sample names as rownames
rownames(loi) <- loi$sample
rownames(tga) <- tga$Name
# Creating new columns for analysis by subtracting crucible mass;
# "dry" (sample mass after drying)
# "c550" (sample mass after 550 C)
# "c950" (sample mass after 950 C)
loi$dry <- (loi$dry_weight - loi$crucible)
loi$c550mass <- (loi$after_550_C - loi$crucible)
loi$c950mass <- (loi$after_950_C - loi$crucible)
loi$c550 <- (loi$dry - loi$c550mass)/(loi$dry) * 100
loi$c950 <- (loi$c550mass - loi$c950mass)/(loi$c550mass) * 100
# Same with TGA
tga$c550 <- (tga$Dry - tga$Mass_550)/(tga$Dry) * 100
tga$c950 <- (tga$Mass_550 - tga$Mass_950)/(tga$Mass_550) * 100Manual LOI results
# Colored boxplots in same graph
loi_long <- loi[c(8,13:14)]
loi_long <- melt(loi_long, id = "context") # Reshaping data to long form
ggplot(loi_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Traditional LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(loi_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Traditional LOI by context",
x ="Context", y = "LOI")
# Color by context
ggplot(loi,
aes(c550, c950, label = rownames(loi), colour = factor(context))) +
geom_point(size=2, aes(shape = factor(type))) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
loi_long = gather(loi, key = var, value = value, c550, c950)
ggplot(loi_long, aes(x = reorder(sample, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
TGA LOI results
# Colored boxplots in same graph
tga_long <- tga[c(8,10:11)]
tga_long <- melt(tga_long, id = "context") # Reshaping data to long form
ggplot(tga_long, aes(x = variable, y = value, color = context)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by temperature",
x ="Temperature", y = "LOI")
ggplot(tga_long, aes(x = context, y = value, color = variable)) +
geom_boxplot() +
labs(title="Thermogravimeter LOI by context",
x ="Context", y = "LOI")
# Color by context biplot
ggplot(tga,
aes(c550, c950, label = rownames(tga), colour = factor(context))) +
geom_point(size=2) +
geom_text_repel(size=3) + labs(title = "Organic vs. carbonate content") +
xlab('Organic LOI (%)') +
ylab('Carbonate LOI (%)') +
theme(axis.title = element_text())
# Stacked barplot (x = reorder(sample, LOI))
tga_long = gather(tga, key = var, value = value, c550, c950)
ggplot(tga_long, aes(x = reorder(Name, value), y = value, fill = var)) +
geom_bar(stat = 'identity', position = 'stack') +
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
Colorimetry
# Libraries
library(dplyr);
library(ggplot2);
library(ggrepel);
library(rgl);
library(plot3D);
library(colordistance)
color1 <- read.table("data/color_AA.txt", header = TRUE, sep = "\t", dec = ",")
rownames(color1) <- color1$Data.Name
color <- subset(color1[, 3:5])
summary(color)## L a b
## Min. :51.37 Min. :3.880 Min. :15.92
## 1st Qu.:57.73 1st Qu.:4.402 1st Qu.:16.49
## Median :58.79 Median :4.555 Median :16.93
## Mean :58.18 Mean :5.067 Mean :17.32
## 3rd Qu.:59.58 3rd Qu.:4.695 3rd Qu.:17.32
## Max. :61.57 Max. :8.670 Max. :21.47ggplot(color,
aes(a, b, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Color (a*b)") +
xlab("a D65") +
ylab("b D65") 
ggplot(color1,
aes(X, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "Luminance") +
ylab("L* (D65)") +
xlab("Sample number") 
ggplot(color,
aes(a, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "A versus luminance") +
ylab("L* (D65)") +
xlab("a") 
ggplot(color,
aes(b, L, label = rownames(color))) +
geom_point(size=2) +
geom_text_repel(size=3) +
labs(title = "B versus luminance") +
ylab("L* (D65)") +
xlab("b") 
# Colorful 3D plot with drop lines
scatter3D(color$a, color$b, color$L,
xlab = "a* (D65)", ylab = "b* (D65)", zlab="L* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$a, color$b, color$L, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)
scatter3D(color$L, color$a, color$b,
xlab = "L* (D65)", ylab = "a* (D65)", zlab="b* (D65)",
phi = 0, bty = "g", type = "h",
ticktype = "detailed", pch = 19, cex = 1.2)
text3D(color$L, color$a, color$b, labels = rownames(color),
add = TRUE, colkey = FALSE, cex = 0.7)